Rough set approach to building expert systems

Knowledge acquisition and uncertain reasoning are crucial in building expert systems. Rough sets theory offers new approaches to acquiring a set of classification rules from a decision table and reasoning under uncertainty. In this paper, a unifying framework based on rough set theory for building an expert system is established. First, an algorithm for rule generation is introduced. Then, several measures of a rule, i.e., support, accuracy, coverage, weight, condition equivalence classes from which the rule is induced, and the length of antecedent, are used to describe the corresponding rule derived from the algorithm. Based on the rules and their measures, some methods of uncertain reasoning are introduced. Examples illustrate the presentation.